Charge neutralization in the active site of the catalytic trimer of aspartate transcarbamoylase promotes diverse structural changes

Abstract

A classical model for allosteric regulation of enzyme activity posits an equilibrium between inactive and active conformations. An alternative view is that allosteric activation is achieved by increasing the potential for conformational changes that are essential for catalysis. In the present study, substitution of a basic residue in the active site of the catalytic (C) trimer of aspartate transcarbamoylase with a non‐polar residue results in large interdomain hinge changes in the three chains of the trimer. One conformation is more open than the chains in both the wild‐type C trimer and the catalytic chains in the holoenzyme, the second is closed similar to the bisubstrate‐analog bound conformation and the third hinge angle is intermediate to the other two. The active‐site 240s loop conformation is very different between the most open and closed chains, and is disordered in the third chain, as in the holoenzyme. We hypothesize that binding of anionic substrates may promote similar structural changes. Further, the ability of the three catalytic chains in the trimer to access the open and closed active‐site conformations simultaneously suggests a cyclic catalytic mechanism, in which at least one of the chains is in an open conformation suitable for substrate binding whereas another chain is closed for catalytic turnover. Based on the many conformations observed for the chains in the isolated catalytic trimer to date, we propose that allosteric activation of the holoenzyme occurs by release of quaternary constraint into an ensemble of active‐site conformations.

Short abstract

PDB Code(s): 5VMQ

Abbreviations

Asp

aspartate

ATCase

aspartate transcarbamoylase

C

catalytic (subunit or trimer)

CbmP

carbamoyl phosphate

NCS

non‐crystallographic symmetry

PALA

N‐phosphonacetyl‐L‐aspartate

WT

wild‐type.

Introduction

Since the pioneering crystallographic studies on yeast hexokinase,1 carboxypeptidase A2 and T4 lysozyme,3 hinge‐bending motions in enzymes have been recognized as important for catalysis. Other examples of hinge closure upon substrate binding include phosphoglycerate kinase,4 gluconate kinase,5 and angiotensin converting enzyme‐related carboxypeptidase, ACE2.6 These conformational changes have been interpreted as supporting the “induced‐fit” theory of enzyme activity.7, 8

An alternative, but not mutually exclusive, theory coined “conformational selection” proposes that enzymes preexist in an ensemble of states, and that substrates preferentially bind to a subset of these enzyme conformations.9 This conformational ensemble can be viewed as analogous to those in the protein folding landscape.10 Experimental evidence for multiple conformational states in the absence of substrates include NMR relaxation studies of the prolyl cis‐trans isomerase cyclophilin A, which show that characteristic conformational changes during catalysis also occur in the free enzyme.11

Conformational changes upon substrate binding also have been observed for many allosteric enzymes such as lactate dehydrogenase,12 phosphofructokinase,13, 14 CTP synthetase15, 16 and aspartate transcarbamoylase (ATCase). In ATCase, the tertiary structural changes include domain closure in the catalytic (C) chains, which is thought to be linked to the quaternary structural changes in the dodecameric enzyme.17 The structures of ATCase holoenzymes in free18, 19 and bisubstrate analog bound states20 display three‐fold crystallographic symmetry and near two‐fold symmetry. The release of structural constraints from the taut‐ (T‐) state holoenzyme previously was proposed to explain the allosteric activation of the enzyme,21 which is consistent with the concerted allosteric model of Monod et al.22 In contrast, the structure of the isolated C trimer exhibits asymmetry in its interdomain hinge conformations.21 The inferred flexibility of the isolated C trimer contrasts with the quaternary constraint of the C trimer when assembled into the holoenzyme. This increased flexibility was hypothesized to explain the higher catalytic activity of the C trimer compared with the relaxed‐ (R‐) state enzyme.21

In the present study, we determined the high‐resolution crystal structure of a mutant C trimer of ATCase, which contains a single amino acid replacement of arginine by alanine (R105A) in the carbamoyl‐phosphate (CbmP) binding site. This substitution decreases the activity of the holoenzyme by 1100‐fold and decreases the apparent affinity for the substrate CbmP by 35‐fold.23 We infer that large tertiary structural changes in the mutant C trimer result from neutralization of a single active site charge and that the opposite conformational changes in the hinge angles of different chains indicate that the tertiary conformations within the trimer are linked energetically.

Results

Structure of the R105A mutant C trimer

We determined the structure of the R105A mutant C trimer to 2.0‐Å resolution by molecular replacement using the structure of the wild‐type C trimer (PDB ID 3CSU)21 as a search model. The structure was refined to a crystallographic R factor of 0.175 and a free R factor of 0.227 with good geometry (Table 1). Most of the residues (897 of 930) in the three polypeptide chains exhibit well‐defined electron density; the ranges of the residues present in the model are provided in Table 2. The site of the R105A substitution has electron density that is consistent with the presence of an alanine side chain and a feature‐enhanced24 omit map confirms the presence of the substitution (Fig. 1).

Table 1.

X‐Ray Data Collection and Refinement Statistics ^a

Crystals
Space group	P212121
Unit cell dimensions (Å)
a	56.03
b	81.10
c	211.65
V m (Å3/Da)	2.34
Data
Resolution range (Å)	18.29–2.01
Unique/measured reflections	57,740/190,953
Completeness (%):	89.8 (82.0)
Average I/σI	11.7 (2.9)
R sym (%)	0.071 (0.353)
Refinement
R cryst	0.175
R free	0.227
RMS Deviations
Bond lengths (Å)	0.007
Bond angles (°)	0.87
Planar groups (Å)	0.005
Φ/Ψ Most Favored (%)	97.5

^avalues in parentheses are for the 2.06 to 2.01‐Å resolution shell.

Table 2.

Interdomain Hinge Angle Comparisons of C Trimer Conformations Compared to the Closed, ATCase‐PALA Conformation

Assembly (sequence, ligand)	PDB ID (Chain ID)	Hinge angle difference (°)a
C trimer (R105A, none)b	5VMQ (A)	14.4
	5VMQ (B)	9.3
	5VMQ (C)	0.2
C trimer (WT, none)21	3CSU (A)	12.0
	3CSU (B)	8.5
	3CSU (C)	6.3
C trimer (WT, PALA)25	1EKX (A)	0.2
Holoenzyme (WT, none)18	6AT1 (A)	5.9

^aThe hinge angle difference (in degrees) is relative to the most closed known conformation (ATCase‐PALA; PDB entry 1D09, chain A)26. Hinge angle differences were calculated as described in Methods.

^bElectron density was observed in chain A for residues 1–77, 85–128, 134–310 (of 310); in chain B for residues 2–231, 247–305; and in chain C for residues 1–310.

Figure 1.

Feature‐enhanced electron density map24 of the active‐site region containing the R105A substitution with residue A105 omitted from the map calculation. The map, which is shown at a contour level of 1.0σ, shows density consistent with an alanine residue at position 105, which is shown in a thicker stick representation. The map was generated using PHENIX27, 28 and the figure was rendered using PyMOL (https://www.pymol.org)

The most distinctive structural feature of the R105A mutant C trimer is the asymmetry in the relative orientations of the two domains within each of the three catalytic chains. Since the tertiary structural differences primarily involve rotation of one domain with respect to the other, the two domains can be viewed as being related by a hinge, with various angular differences between conformations. After superimposing one of the two structural domains of each chain, large differences are seen in the relative positions of the other domain among the three chains [Fig. 2(A)], which differ by as much as 14.4° (Table 2).

Figure 2.

Hinge differences among chains of the R105A mutant C trimer. The chains were superimposed based on N‐terminal domain residues 1–73 and 90–134 and the hinge differences can be seen from the relative displacement of C‐terminal domains. A, Comparison of the three chains in the R105A mutant C trimer. The chains are shown as Cα traces; chain A, green; B, red; and C, violet. B, Comparison of the R105A mutant C trimer (chain B, red) to the wild‐type C trimer (chain B, light blue)21 shows the intermediate hinge conformations of the two structures. The 240s loop, which is disordered in both chains, is not depicted. C, Comparison of the most closed chain conformation of the R105A mutant C trimer (chain C, violet) with a chain from the C trimer‐PALA complex (1EKX)25 (chain C, light green)

Two active‐site loops, the 80s and 130s loops, exhibit low electron density in one of the chains (A), and the 240s loop displays low density in another chain (B), which indicates that their conformations are disordered [represented by gaps in the Cα traces in Fig. 2(A)]. For the active‐site loops that exhibit well‐defined electron density, their conformations are variable. A noteworthy example is the 240s loop, which is reoriented by 21° in chain C relative to chain A [Fig. 2(A)]. Progressively smaller differences in loop conformations are evident in the 130s, 80s, and 50s loops [Fig. 2(A)]. Following superposition of the amino‐terminal domain of each chain, a plot of the difference in the alpha carbon (Cα) position for each residue depicts the relative displacement of the carboxyl‐terminal domain, as well as the differences in local conformations for the three chains [Fig. 3(A)].

Figure 3.

Hinge differences or similarities as shown from plots of pairwise Cα differences. The chains were superimposed based on N‐terminal domain residues 1–73 and 90–134 and the hinge differences can be seen from the relative displacement of C‐terminal domains. A, Pairwise differences among the three chains in the R105A mutant C trimer. Chain A vs. B, black symbols; chain A vs. C, dark grey symbols; chain B vs. C, light grey symbols. Comparisons correspond to the superpositions shown in Figure 2(A). B, Pairwise differences between the R105A mutant C trimer (chain B) and WT C trimer (chain B)21, black symbols, and between the R105A mutant C trimer (chain C) and C trimer‐PALA (chain C)25, grey symbols. Comparisons correspond to the superpositions shown in Figure 2(B,C)

Comparison to wild‐type C trimer and holoenzyme structures

The structure of the R105A mutant C trimer is asymmetric in its hinge conformations, with differences of up to 14.4°. These differences are considerably greater than those observed in the wild‐type C trimer, which differ by up to 5.7°.21 One of the chains in the R105A mutant C trimer (chain A) is 2.4° more open than the most open chain (A) of the wild‐type C trimer (Table 2). The second (chain B) is slightly more open (0.8°) than the intermediate conformation (chain B) of the wild‐type C trimer. The most closed chain of the R105A mutant C trimer (chain C) is 6.1° more closed than the corresponding chain of the C trimer (C). The hinge conformation of chain C of the R105A mutant trimer is nearly identical (0.2° more open) to all three chains in the structure of the C trimer in a complex with the bisubstrate analog PALA25 (Table 2). This result is surprising since the bisubstrate analog tethers the two domains of each chain in a closed conformation in the C trimer‐PALA complex.25 Further, all of the previously determined ATCase structures without active‐site ligands (e.g., 18, 19, 21) display relatively open hinge conformations.

Like the wild‐type C trimer, the R105A mutant C trimer contains disordered active‐site loops. However, the locations of the disordered loops, and the number of each, differ between the two structures. For example, the 80s loop is disordered in one chain (A) of the R105A mutant C trimer, whereas this loop is disordered in all three chains of the wild‐type C trimer21. The 130s loop is disordered in one chain (A) of the R105A trimer, but is ordered in all three chains of the wild‐type C trimer. The 240s loop is disordered in one chain (B) of the R105A trimer, compared with two chains in the wild‐type C trimer. In contrast to the wild‐type and mutant C trimers, all of the active‐site loops in the C trimer‐PALA complex are ordered because these loops directly contact the bisubstrate analog. In all of the known C trimer structures, the 240s loop either is disordered or adopts one of two ordered conformations, which differ in orientation relative to the remainder of the carboxyl‐terminal domain by 21° [Fig. 2(A)]. Two of the three chains of the R105A mutant C trimer exhibit significant differences in loop positions compared to both the wild‐type C trimer and its complex with PALA.

The third chain (C) of the R105A mutant C trimer, which resembles the closed conformation of the C trimer‐PALA complex in its interdomain conformation, also resembles the latter in several of its active‐site loop conformations [Fig. 2(C)]. Whereas the 240s loop of chain C is quite similar to that in the C trimer‐PALA complex, the 80s loops differ by as much as 5 Å in their respective Cα positions [Fig. 3(B)]. Most of the active‐site residues adopt conformations that are similar to those in the presence of bound PALA (Fig. 4). A noteworthy exception is the side chain of Arg54, which occupies a position that sterically occludes the active site. The side chain of Arg54 makes several hydrogen bonds with neighboring residues; Arg54 NH1 and NH2 are positioned 3.0 and 3.5 Å, respectively, from the carbonyl oxygen of Leu267, and Arg54 NH2 is 3.0 Å from Gln137 NE2. The interaction between Arg54 and Leu267, which are in different structural domains, appears to stabilize the closed hinge conformation in the absence of bound substrates or analogs.

Figure 4.

Differences in active site residues in the most closed chain of the R105A mutant C trimer (chain C, violet) and a chain from the C trimer‐PALA complex (chain C, light green) (1EKX).25 Chains were superimposed using the backbone atoms of all residues. Known substrate binding residues are shown. Residues S80 and K84 (denoted with prime (′) are contributed by an adjacent chain (A) and are disordered in this chain of the R105 mutant C trimer. The bisubstrate analog from the C trimer‐PALA complex is shown at the center in dark gray. The side chain of Arg54 in chain C of the R105A mutant occupies the PALA‐binding site and mediates domain‐bridging interactions, which stabilizes the closed conformation in the absence of substrates

The structure of the R105A mutant ATCase holoenzyme (PDB ID 1I5O)29 closely resembles the structure of non‐liganded ATCase (PDB ID 6AT1)18, except at the site of the residue substitution. Therefore, we compared the structure of the R105A mutant C trimer to non‐liganded ATCase, whose structure had been determined to a higher resolution limit than the R105A mutant holoenzyme. The hinge conformations of the R105A mutant C trimer range from 9° more open (chain A) to 6° more closed (chain C) than the catalytic chains in the free ATCase holoenzyme. The active‐site loops that are disordered in the R105A mutant C trimer are ordered in the wild‐type and R105A holoenzymes due to interchain contacts not present in the C trimer. Specifically, in the holoenzyme, the 130s loop contacts a catalytic chain from the apposing trimer, the 240s loop contacts apposing catalytic and regulatory chains and the 80s loop contacts an adjacent catalytic chain. Since the C trimers in the ATCase‐PALA complex (6AT1)18 and (1D09)26 closely resemble the C trimer‐PALA complex (1EKX)25 described above, separate comparisons between the R105A mutant C trimer and the ATCase‐PALA complex did not yield any additional insights.

Discussion

It is noteworthy that a T4 lysozyme mutant with a substitution between the two structural domains previously was observed in five different conformations, with interdomain hinge displacements of up to 32°.3 In our previous study, the wild‐type ATCase catalytic trimer exhibited three different conformations21 and a fourth conformation was seen in the presence of a bisubstrate analog,25 with hinge differences as large as 12°. In the present study, we observe two new conformations in an ATCase catalytic trimer mutant, which possesses a substitution, R105A, in the active‐site cleft between the two domains. For both lysozyme and C trimer mutants, the observations of multiple conformations are interpreted to represent flexibility of the enzymes in solution.

In the structure of the wild‐type C trimer,21 R105 mediates polar interactions with the side chains of residues E50 and S52, and with the backbone carbonyl oxygen atoms of A127 and G130. Since R105 is in the amino‐terminal domain and does not contact residues in the carboxyl‐terminal domain (residues 150–284, see Methods), we infer that the large and diverse structural changes in the R105A mutant C trimer are attributable to partial neutralization of positive charge in the active site. Collectively, the R105A mutant C trimer and the previously described structure of the wild‐type C trimer exhibit six different interdomain hinge conformations. This structural variation suggests that the C trimer is flexible in solution21 and that charge neutralization in the active site, either by mutagenesis as described here, or naturally by deprotonation of basic residues, binding of substrates or other ions in the active sites can lead to large conformational changes.

It is striking that the same substitution led to opposite conformational changes in the hinge angle between domains in two of the chains in the mutant C trimer relative to the wild‐type C trimer. This observation suggests that the conformational changes are structurally and energetically linked. The asymmetry in the two structures and the opposing hinge changes in R105A mutant C trimers may explain previous observations of functional heterogeneity in the active sites of the wild‐type C trimer with respect to binding the first substrate CbmP30. Moreover, the bisubstrate analog PALA bound to the wild‐type C trimer with a stoichiometry of 2.3 PALA per C trimer in two independent studies.31, 32 Although the high‐resolution crystal structure of the C trimer in a complex with PALA shows three molecules of PALA bound with full occupancy25, another crystal structure was obtained with two PALA molecules bound and all three hinges were in the closed conformation (authors' unpublished data). Therefore, the stoichiometry of 2.3 PALA bound per C trimer suggests that either two or three PALA can bind the wild‐type C trimer and that binding of the third PALA molecule might depend on the kinetics of the associated conformational changes.

Interestingly, in an earlier study, hybrid ATCase holoenzymes containing the R105A substitution in five of the six catalytic chains were used to assess the structural effects of PALA binding to one of the six catalytic chains by small‐angle X‐ray scattering.29 Surprisingly, the conformational changes seen in the R105A mutant C trimer were not seen in the crystal structure of the R105A holoenzyme, which resembled the free, T‐state holoenzyme29. This result reinforces the earlier proposal that the assembly of C trimers into holoenzyme imposes symmetry21.

In summary, the marked asymmetry of the R105A mutant C trimer, in contrast to the moderate asymmetry of the wild‐type C trimer, shows that large conformational changes likely result from neutralization of a single active‐site charge. The potential for large conformational changes promoted by neutralization of an active‐site charge suggests that the catalytic chains likely access multiple conformations during a catalytic cycle, including open conformations that allow substrate binding and/or product release, and a closed conformation for transition‐state stabilization. Furthermore, the potential for these conformations to be coupled suggests that catalytic turnover may occur among the three active sites in an alternating, or cyclic, fashion. Although the potential for conformational changes is likely to be lower in the more constrained holoenzyme21, catalysis in ATCase likely depends on similar sampling of multiple conformational states during a catalytic cycle. The proposal that allosteric activation in the ATCase holoenzyme occurs through the release of structural constraints into an ensemble of conformations will require further structural studies, including those on mutant ATCase enzymes whose equilibrium favors the activated R‐state.

Materials and Methods

Site‐specific mutagenesis

The R105A substitution was introduced into the E. coli pyrB gene by PCR mutagenesis.33 Two overlapping fragments were generated, each using one primer that flanked the gene and one mutagenic primer. The forward and reverse mutagenic oligonucleotide sequences were: 5′ CGATAGTGATGGCTCATCCGCAGG and 5′ CCTGCGGATGAGCCATCACTATCG, respectively. The overlapping fragments were mixed in an equimolar ratio and the mutant pyrB gene was amplified using the flanking primers. The PCR product was digested with the restriction endonucleases NcoI and HindIII and ligated into the ATCase expression plasmid pAX4, which was derived from pRCP000.34

Protein purification and crystallization

ATCase containing the R105A substitution in the catalytic chains and a His₆ sequence at the N‐terminus of each regulatory chain was purified as previously described for ATCase His₆ enzyme.34 Mutant C subunits were isolated from the His₆‐tagged holoenzyme by treatment with neohydrin followed by ion exchange chromatography.35 Orthorhombic crystals of the C_R105A mutant trimer were obtained by vapor diffusion at 4°C from hanging drops containing 5 µL 0.1 M Tris‐HCl, pH 7.8, 0.04 M CaOAc, 7.5% PEG 8000 and 5 µL of C subunit (6.1 mg/mL) in 10 mM Tris‐HCl, pH 7.5, 1 mM 2‐mercaptoethanol. The crystals grew in space group P2₁2₁2₁, with unit cell dimensions a = 56.03, b = 81.10, c = 211.65 Å (Table 1).

Structure determination and refinement

Crystals (ca. 1.0 × 0.3 × 0.2 mm) were equilibrated to 40 mM Tris^.Cl, pH 7.5, 40 mM CaOAc, 29% PEG 8000, 22% methylpentanediol and flash cooled on a 0.1 mm nylon loop (Hampton Research) in liquid N_2. The X‐ray data were collected from a single crystal at 100 K using a Mar345 image plate on beam line 7–1 at the Stanford Synchrotron Radiation Laboratory. The data were integrated and scaled using MOSFLM36 and SCALA37 to give R _sym = 0.071 for 57,740 independent reflections from 20.0–2.0‐Å resolution. The data were 90% complete overall, with average I/σI values of 11.7 (2.9 in the 2.06–2.01 Å shell; Table 1). Molecular replacement was carried out with AMoRe38 using the structure of the wild‐type C trimer21 as a search model.

Rigid‐body refinement of individual domains was followed by positional and restrained B factor refinement using TNT.39 Further stages of refinement were performed with refmac40 and phenix.refine27 using three‐fold non‐crystallographic symmetry (NCS) restraints for each of the two structural domains. Model building was carried out using XtalView41 and Coot.42 Iterations of refinement and manual building were carried out until the R _free converged. The NCS restraints were relaxed at the end of refinement, which resulted in a slight decrease in R _free. The final model contains 6988 protein atoms in three polypeptide chains, 429 water molecules, one Ca²⁺ and one Cl^‐ ion. The structure has been refined to a crystallographic R factor of 0.175, a free R factor of 0.227 with good geometry (Table 1). Ninety‐eight percent of residues are in the favored regions of a Ramachandran plot and only the arginine at position 270 occupies a disallowed region, as has been observed in other high‐resolution ATCase structures.21, 26 The atomic coordinates and structure factors have been deposited in the Protein Data Bank (http://www.rcsb.org; accession number 5VMQ).

Structural analysis

Superpositions and hinge angle calculations were performed using GEM.43 Hinge angle differences were calculated as the rotation angle required to bring an amino‐terminally superimposed pair of molecules into carboxyl‐terminal superposition. Superpositions were based on residues comprising the amino‐terminal (1–73 and 90–134) or carboxyl‐terminal (150–229 and 252–284) domains and excluded the two interdomain helices (135–149 and 285–305) and two variable active‐site loops (residues 74–89 and 230–251). Active‐site comparisons were done based on backbone atom superpositions of entire chains. Structural superpositions and figures were generated using PyMOL (https://www.pymol.org).